This repository is for the development of a C++ extension for a pythonic implementation of the radial fingerprints used in the Rapid Artificial Neural Network of Dr. Kip Barrett and Dr. Doyl Dickel's work here: https://www.sciencedirect.com/science/article/abs/pii/S0927025621002068

Simple pair interactions are considered and summed over all neighbors of a given atom. The pair-interaction structural fingerprint for atom $\alpha$ is defined as:

where $\beta$ labels neighboring atoms within the cutoff radius of atom $\alpha$.

The cutoff radius $r_c$ is defined using a MEAM-based piecewise function:

The cutoff smoothly transitions from 1 to 0 to account for negligible contributions from atoms at large separations. This ensures locality while maintaining differentiability for learning-based interatomic potentials.